Kava is NOT GABA-active in a meaningful way, if at all.
I think you may be referring to the fact that flumanezil doesn't appear to block the anxiolytic effects of kava exctracts as was shown in this paper:
Kava extracts produce significant murine anxiolytic-like behavioral changes and sedation that are not mediated through the benzodiazepine binding site on the GABA(A) receptor complex.
pubmed.ncbi.nlm.nih.gov
So there are a few different binding sites for the GABAa receptor and GABA, benzodiazepines, and barbiturates all bind at different sites. Flumazenil specifically competitively blocks the benzo site but only in a neutralizing manner, it doesn't appear to decrease the efficacy of endogenous GABA binding
There was a recent paper where experiments were performed on a variety of human GABAa receptors expressed in oocytes, they found that kavain has a modulatory effect similar to barbiturates and that it may bind at a site similar to some anesthetics.
"We demonstrate that the modulatory effect of kavain at GABAARs occurs in a subtype non-selective and flumazenil-insensitive manner. We also present evidence, for the first time, that kavain action at GABAARs is attenuated by the anaesthetic-impairing β3N265M point mutation."
"We investigated the effect of combining benzodiazepine site ligands with kavain at α1β2γ2L GABAARs (
n = 5;
Fig 4). In concordance with the literature, diazepam (1 μM) enhanced current responses elicited by 10 μM GABA in a flumazenil-sensitive manner, and flumazenil alone did not alter GABA responses, consistent with its neutralising modulator profile [
33]. Unlike diazepam, the GABA-enhancing action of 300 μM kavain was unaffected in the presence of flumazenil. We also found that the co-application of kavain and diazepam resulted in significantly greater enhancement of GABA current (350 ± 10%) compared to their actions alone (GABA + diazepam: 260 ± 4.2%,
p < 0.001; GABA + kavain: 260 ± 11%,
p < 0.0001; paired
t test;
Fig 4B). However, this degree of potentiation was less than additive (expected additive value = 420%). The same experiment was repeated with 100 nM diazepam, and a less-than-additive interaction was detected again with the kava and diazepam combination (
n = 3; data not shown)."
"Next, we assessed the possibility of kavain interacting with the transmembrane β+α‒ anaesthetic binding sites. To this end, we introduced the α1M236W, β2M286W and β3N265M point mutations which perturb etomidate and propofol actions, and investigated the impact of these mutations on kavain activity at GABAARs."
"Similar differential enhancement effects that arise from the distinct GABA efficacies at synaptic αβγ and extrasynaptic αβδ receptors have been described for other allosteric ligands such as etomidate [
38], propofol [
39], pentobarbital [
40], and neurosteroids [
41]. While our findings do not conclusively prove that GABAARs are the
in vivo substrate for kavain, it is tempting to speculate that kavain may have a larger physiological impact on extrasynaptic than synaptic GABAARs."
"The low affinity of kavain detected in our study is in agreement with most studies conducted in the past using kava extracts or pure kavalactones [
21,
22,
46]. The high-micromolar concentrations required to observe the pharmacological actions of kavalactones
in vitro raise the critical question of whether the concentrations used in these studies are physiologically relevant. Studies conducted in mice showed that 100 mg/kg of kavain alone caused a rapid surge in brain concentration (up to 100 μM) within minutes after
i.
p. injection [
22,
47]. A higher brain concentration of kavain was found (120 μM) in the presence of other kavalactones (44 mg kavain in 120 mg/kg kavalactone mixture;
i.
p.), clearly demonstrating pharmacokinetic synergism [
22,
47]. Davies
et al. (1992) predicted higher brain concentrations (300 μM) with the administration of larger doses (300 mg/kg) of kavain [
22]. Thus, kavain concentrations used in our study appear to correlate well with the concentrations needed to induce anxiolysis and hypnosis in mice [
27,
48]. Unfortunately, the relevance of this concentration range in humans cannot be established as no data is currently available [
49]."
Extracts of the pepper plant kava (Piper methysticum) are effective in alleviating anxiety in clinical trials. Despite the long-standing therapeutic interest in kava, the molecular target(s) of the pharmacologically active constituents, kavalactones have not been established. γ-Aminobutyric acid...
journals.plos.org
It is still relatively unknown if the primary anxiolytic effects are due to the specific GABAa modulation investigated in this paper and there are other studies which investigate inhibitory modulation through sodium and calcium channels as well. Like other phytochemicals, kavalactones are quite promiscuous and bind at a variety of receptors but GABAa interactions may very well play a role.